A Novel Substrate-Binding Site in the X-ray Structure of an Oxidized E. coli Glyceraldehyde 3-Phosphate Dehydrogenase Elucidated by Single-Wavelength Anomalous Dispersion

Escherichia coli (E. coli), one of the most frequently used host for the expression of recombinant proteins, is often affected by the toxic effect of the exogenous proteins that is required to express. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) is a multi-functional protein that has been used as a control marker for basal function and it is known to undergo cysteine oxidation under different types of cellular stress. Here, we report the 3D structure of the endogenous GAPDH purified from stressed E. coli cells expressing a eukaryotic protein. The structure was solved at 1.64 Å using single-wavelength anomalous dispersion (SAD) phasing with a selenium-modified enzyme. Interestingly, each GAPDH monomer contains a molecule of glyceraldehyde-3 phosphate in a non-previously identified site. Furthermore, the catalytic Cys149 is covalently attached to a ~300 Da molecule, possibly glutathione. This modification alters the conformation of an adjacent alpha helix in the catalytic domain, right opposite to the NAD+ binding site. The conformation of the alpha helix is stabilized after soaking the crystals with NAD+. These results exemplify the effects that the overexpression of an exogenous protein has over the host proteins and sheds light on the structural changes that large oxidant molecules on the catalytic cysteine produce for the GAPDH enzyme.


Introduction
Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) is a ubiquitous, highly studied enzyme that reversibly transforms glyceraldehyde 3-phosphate to 1,3-diphosphoglycerate in the presence of NAD + and inorganic phosphate in the glycolysis pathway [1].Due to its central participation in glycolysis and its high abundance, it is considered a biological marker of basal activity [2,3].Nonetheless, apart from its glycolytic function, GAPDH displays a battery of moonlighting activities.In eukaryotic organisms, it participates in membrane transport, DNA repair and replication, autophagy, pathogenesis, host immunosuppression, and microtubule building, among others [4][5][6][7][8].In bacteria, several protein-binding partners have been identified for GAPDH, suggesting its participation in a variety of non-canonical functions, similar to the ones described in eukaryotes [9].Bacterial GAPDH is known to act as a virulence factor in pathogens [10][11][12], it is secreted in probiotics like Lactobacillus plantarum [13], and participates in DNA repair, being one component of a complex of proteins dedicated to this end [14].
EcGAPDH was purified from BL21(DE3) E. coli cells (Merck Millipore, KGaA, Darmstadt, Germany) transformed with a pET28a (+) vector, carrying a synthetic E. coli codon optimized sequence of a human gene that codifies for a 32.4 kDa protein (GeneScript, New Jersey, US).An initial overnight culture containing 100 µg/mL ampicillin (Sigma-Aldrich, St. Louis, Mo, US) was grown during 16 h at 37 • C with constant shaking at 200 rpm.The cell pellet from this culture was recovered by centrifugation for 5 min at 4500 g and resuspended in 1 L of sterile LB-Miller (Sigma-Aldrich, St. Louis, Mo, US) or homemade M9 media.To prepare the M9 media, we dissolved 64 g of Na 2 HPO 4 7•H 2 O, 15 g KH 2 PO 4 , 2.5 g of NaCl and 5 g of NH 4 Cl in 1 L of distilled H 2 O and sterilized (M9 salts).Then, 200 mL of the sterile M9 salts and 2 mL of filtered-sterilized 1M MgSO 4 , 20 mL of 20% glucose and 100 µl of 1M CaCl 2 were added, and the volume was brought out to 1 L with distilled H 2 O.
The cultures were grown at 37 • C in constant agitation at 200 rpm until the optical density at 600 nm reached 0.4.Then, for the LB-miller media, we induced the expression of the human protein cloned in the pET28a (+) vector (HACD1 gene) by adding 1mM Isopropyl βd-1-thiogalactopyranoside (IPTG).In the case of the minimal media, 50 mg of selenomethionine, 100 mg of lysine, 100 mg threonine, 100 mg phenylalanine, 50 mg leucine, 50 mg isoleucine and 50 mg of valine were added, and after 15 min of incubation with the additives, we induced protein expression with 1 mM IPTG.Both cultures were allowed to grow during 14 h at 18 • C, and cell pellets from the cultures were recovered by centrifugation at 6400 g.

Protein Purification
Cell pellets were resuspended in lysis buffer containing 20 mM Na 2 HPO 4 , 50 mM NaCl, one tablet of complete protease (Roche Diagnostics, Indianapolis, IN, USA) and 1 mM PMSF (Amresco, Solon, OH, USA) at pH 7.2.The cell suspension was disrupted by sonication for eight minutes on ice, and the lysate was centrifuged 45 min at 20,000 g.The supernatant was filtered two times before chromatography, first with a 0.45 µm PVDF membrane and then with a 0.22 µm PVDF membrane (Merck Millipore, KGaA, Darmstadt, Germany).The filtered lysate was loaded onto a hydroxyapatite column equilibrated in with 20 mM Na-phosphate buffer pH 6.8, the column was washed using the same buffer and the protein was eluted with a linear gradient from 20 mM to 500 mM Na-phosphate buffer pH 6.8 and 50 mM NaCl.The fractions containing more protein were passed through an affinity His prep FF 16/10 column pre-equilibrated with 5 mM imidazole and 50 mM NaCl, washes contained 50 mM NaCl and 50, 100, or 250 mM imidazole each, the protein eluted at 100 mM imidazole.After affinity chromatography, the protein was concentrated and loaded onto a molecular filtration S100 column equilibrated in 10 mM HEPES pH 6.8 and 20 mM NaCl.Fractions containing the purest protein were concentrated and used for dynamic light scattering (DLS), circular dichroism (CD), and crystallization experiments.

Crystallization
Initial crystallization conditions were carried out with crystal screens I and II (Hampton Research Aliso Viejo, CA, US) and Nextal PACT Suite (Qiagen, Germantown, MD, US) with the native, non-modified enzyme at a final concentration of 5.5 mg/mL in 10 mM HEPES pH 7.2.The crystallization trays were incubated at 18 • C.After six days of incubation, several conditions contained crystals, but only two of these resulted promising D2 from CSI and H11 from PACT the Suite.The best crystals were obtained in the optimized conditions from CSI, contained 0.1 M HEPES pH 7.5 and 1.43 M sodium citrate, and grew after three days of incubation.The selenium-modified protein was crystallized using only the optimized crystallization conditions identical to those of the native enzyme.

Soaking Experiments
Crystals from the selenium modified enzyme grown in the optimum crystallization conditions, were soaked for 10 min in 5 mM NAD + dissolved in mother liquor; one crystal was then cryo-protected with mother liquor and 25% trehalose and collected in the in-house X-ray diffractometer at the Laboratorio Nacional de Estructura de Macromoléculas (LANEM-IQ-UNAM).

Data Collection and Processing
The two Se-modified protein crystal were cryo-protected with 25% trehalose, 0.1 M HEPES pH 7.5, and 1.43 M sodium citrate.In the case of the native protein, the cryo-protectant solution contained 30% glycerol and all the other components of the mother liquor.All crystals were frozen directly in the cryo-stream and used to collect a complete dataset.For the Se-modified crystals, one data set was collected in order to obtain high redundancy for SAD phasing.Data collection was performed at LANEM-IQ-UNAM with a Rigaku MicroMax-007HF rotating anode and a Dectris Pilatus3R 200K-A detector (Rigaku, The Woodlands, TX, US).Data were processed using HKL3000 [26], and the spatial group was revised with the program Pointless from CCP4 [27].The .mtz file from pointless was then used in AutoSol (Phenix) [28] to obtain the phases.Initial modeling was done automatically with ArpWarp (CCP4) [27], followed by manual modeling using Coot [29].

Model Refinement
The quality of the diffraction data and the presence of Se in the protein allowed us to obtain a model using SAD.AutoSol (Phenix) gave an initial result for the phase.Manual modeling allowed us to finish the construction of one monomer, and the sequence of the model was blasted (Protein-BLAST) to determine the identity of the protein.GAPDH from E. coli was the main hit.Refinement was continued, and we discovered the presence of three main positive, unknown electron density blobs.We used LigandFit from Phenix to identify the ligands, which were trehalose (TRE) and glyceraldehyde 3-phosphate (G3P).Glutathione (GSH) was manually built; however, despite our attempts, we could not achieve a good fit to the electron density map.All ligands were built using coot and the .CIF files from the CCP4 library.To obtain a model with two conformations in an entire alpha helix, the last steps of model refinement were done using two individual models, for each model one of the two observed conformations for residues 206-214 (Ser, Ser, Thr, Gly, Ala, Ala, Lys, Ala, Val) were used.The coordinates for the residues comprising this alpha helix were merged in a single model, given occupancy of 0.5 and refined.To obtain the GAPDH models for the native protein, and from the soaking experiment, we used the final model from the Se-modified protein during molecular replacement (MR) using Phaser-MR from Phenix [30].Removal of all solvent molecules, ligands, and the alpha helix comprised by 206-214 was performed before MR.The models were refined using Phenix.refine[31], and the missing residues were added manually according to the observed electron density using Coot [29].Final models were validated and deposited in the PDB database (6UTO, 6UTN, and 6UTM).

PDB Search and Comparison
The Protein Data Bank was used to find all the available structures of GAPDH.Employing the sequence from the E. coli enzyme, an initial hit of 171 entries was obtained.From these, manual inspection allowed us to remove 11 structures, which corresponded to other enzymes, then a set of 160 models was analyzed.The presence of ligands and their location in the structure was obtained from visual inspection using Coot and from the PDB entries.

Matrix-Assisted Laser Desorption/Ionization-Time of Flight (MALDI-TOF)
We determined the mass of EcGAPDH from the protein contained in three crystals from the native and three crystals from the Se-modified protein, separately.Sinapinic acid was used as matrix and was added directly to the crystals, which were loaded onto a MALDI plate.Matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) was performed in a Microflex Bruker instrument (Bruker, Billerica, MA, US), with a nitrogen laser at a wavelength of 337 nm, with positive polarity and acceleration of 20 kV using the linear acquisition operation.Here, 1000 shots were used to collect the spectrums.Bovine serum albumin (66,431 Da) and a chitinase from agave (31,742.10Da) [32] were used as calibration markers.

Crystals
After six days, crystals grew to maximum dimensions of 300 µm × 200 µm.Characterization of the crystals showed that they diffracted X-rays to a maximum of 1.64 Å and belonged to the space group C2221, with unit cell dimensions of a = 77.273,b = 186.709,c = 122.132,and α = β = γ = 90 (Table 1).All crystals contained two GAPDH monomers in the asymmetric unit with a Matthews' coefficient of 2.08 Å 3 Da −1 and a solvent content of 40.80%.The GAPDH dimer forms a tetramer with symmetry-related molecules.Three crystallographic structures were solved-one for a Se-modified enzyme, one for the native enzyme, and one for the Se-modified enzyme soaked in NAD + .

Identification of GAPDH
For this study, our initial intention was to express a human protein with a molecular mass of 32.4 kDa in E. coli cells (see methods).Instead, we purified an endogenous protein from the bacteria, which we determined to be homogeneous using DLS, and to be folded with a CD spectrum in the far-UV region (data not shown).Therefore, we conducted the crystallization experiments, and suitable crystals grew after six days at 18 • C. The selenium modified protein was prepared and crystallized to obtain the experimental phases.However, when we analyzed the partial sequence of the model built with ArpWarp and Coot, it did not match the sequence of the initial target protein; thus, suggesting the presence of a different protein in the crystals.A MALDI-TOF spectrum was obtained using three crystals from the native, non-modified protein, and the mass corresponded to 35.5 kDa (Figure 1).Finally, the complete protein sequence and its identity were obtained from the experimental model obtained using SAD and BLAST (NCBI), respectively.

GAPDH Structures
Once that we identified GAPDH in the Se-modified crystal and obtained a refined model, we determined the three-dimensional structure of EcGAPDH from the native and NAD + crystals.Overall, the structure of GAPDH is considerably similar in the three crystals.Superposition between the native and Se-modified crystals gives an RMSD of 0.16 Å, whereas the RMSD between the Se-modified model and the Se-modified crystal soaked in NAD + is 0.15 Å.Thus, there are no global differences in the overall structures.In the three cases, there is a dimer in the asymmetric unit, and GAPDH Crystals 2019, 9, 622 6 of 14 forms a crystallographic tetramer with symmetry related molecules (Figure 2a).Two domains compose EcGAPDH, the NAD + binding domain, residues 1-148, 312-330 (E. coli numbering) and the catalytic domain, residues 148-311 [1,33,34].The NAD + binding domain consist of six parallel and one anti-parallel β-strands, surrounded by four α-helices, the typical structure of a Rossmann fold.Residues from two subunits form the NAD + binding site, in one monomer residues Gly10, Arg11, Ile12, Asp33, and Asn313 have close contacts with the NAD + molecule, nonetheless residues Gly187 and Pro188 from the adjacent monomer are near the cofactor (~4.2-5.0Å).A sheet of seven anti-parallel β-strands and four α-helices (residues 149-165, 209-216, 250-265, and 313-330) form the catalytic domain.At the beginning of the first α-helix, residues Cys149 and His176 constitute the catalytic site in EcGAPDH.Helix 149-165 is adjacent to the α-helix formed by residues 209-216 and both the catalytic Cys149 and the later helix are right opposite to the NAD + binding region (Figure 2b).

GAPDH Structures
Once that we identified GAPDH in the Se-modified crystal and obtained a refined model, we determined the three-dimensional structure of EcGAPDH from the native and NAD + crystals.Overall, the structure of GAPDH is considerably similar in the three crystals.Superposition between the native and Se-modified crystals gives an RMSD of 0.16 Å, whereas the RMSD between the Semodified model and the Se-modified crystal soaked in NAD + is 0.15 Å.Thus, there are no global differences in the overall structures.In the three cases, there is a dimer in the asymmetric unit, and GAPDH forms a crystallographic tetramer with symmetry related molecules (Figure 2a).Two domains compose EcGAPDH, the NAD + binding domain, residues 1-148, 312-330 (E. coli numbering) and the catalytic domain, residues 148-311 [1,33,34].The NAD + binding domain consist of six parallel and one anti-parallel β-strands, surrounded by four α-helices, the typical structure of a Rossmann fold.Residues from two subunits form the NAD + binding site, in one monomer residues Gly10, Arg11, Ile12, Asp33, and Asn313 have close contacts with the NAD + molecule, nonetheless residues Gly187 and Pro188 from the adjacent monomer are near the cofactor (∼4.

Oxidized Cys149 in the GAPDH Model
A quick analysis of the electron density maps in the Se-modified crystal revealed several interesting sites with bulky positive electron density in the Fo-Fc map.Notably, a large blob that seemed to be bound to the catalytic Cys149 caught our attention (Figure 3a).After the first rounds of refinement, an initial search for a possible transition state intermediate was unsuccessful; therefore, we searched for other molecules that could oxidize the enzyme.Since the positive density was large,

Oxidized Cys149 in the GAPDH Model
A quick analysis of the electron density maps in the Se-modified crystal revealed several interesting sites with bulky positive electron density in the Fo-Fc map.Notably, a large blob that seemed to be bound to the catalytic Cys149 caught our attention (Figure 3a).After the first rounds of refinement, an initial search for a possible transition state intermediate was unsuccessful; therefore, we searched for other molecules that could oxidize the enzyme.Since the positive density was large, one of the best possible explanations to occupy this density was glutathione (GSH) [35,36].The molecule was manually modeled with the GSH.CIF file from CCP4.J ligand (coot) was used to build sections of the GSH that better fitted to the density map, and finally, the entire GSH was modeled.The use of GSH gave the best R free and R work values for the model; nevertheless, despite our attempts several clashes between the GSH and the protein were present (Figure S1), thus we decided to remove the GSH from the final model (6UTO).Therefore, a large positive electron density is observed in the Fo-Fc maps (Figure 3a,b).Positive electron density for the modification near Cys149 was observed in two of the three structures obtained in this study, those with the Se-modified protein and the native, non-modified protein.However, the Fo-Fc map electron density for the native crystal was weaker than the one observed for the Se-modified protein.Only the crystals soaked in NAD + showed a significant reduction in the electron density surrounding Cys149; in this case, the residue was modeled as a cysteine sufinic acid (Figure 4b).These results suggest that the presence of the exogenic plasmid or the expression of the human protein could have produced the oxidation of the Cys149; however, the stress caused by the use of minimum media, increased the number of modified EcGAPDH, this data is in agreement with the higher amount of stress observed in E. coli during the expression of recombinant proteins [37].To characterize the identity of the modification in Cys149, we performed MALDI-TOF experiments.The native non-modified protein crystals displayed a mass of 35,506.7 Da, and for the Se-modified protein crystals a mass of 36,142.0Da was obtained, a difference of 635.3 Da is observed, suggesting the presence of one GSH and seven Se-methionines per GAPDH monomer (Figure 1).

Double Conformation of the Helix (Residues 206-214)
During model refinement, some residues of the alpha-helix adjacent to the catalytic site (res.204-214) displayed positive electron density in their surroundings.The positive density suggested the existence of two conformations for these residues; phenomenon that only occurred when the catalytic Cys149 exhibits a large electron density in its surroundings (native and Se-modified models, PDBs 6UTM, 6UTO).Occupancies for one or the other conformer for each of the nine residues refined to values of 0.59 and 0.41, thus supporting the presence of the two simultaneous conformers.This

Double Conformation of the Helix (Residues 206-214)
During model refinement, some residues of the alpha-helix adjacent to the catalytic site (res.204-214) displayed positive electron density in their surroundings.The positive density suggested the existence of two conformations for these residues; phenomenon that only occurred when the catalytic Cys149 exhibits a large electron density in its surroundings (native and Se-modified models, PDBs 6UTM, 6UTO).Occupancies for one or the other conformer for each of the nine residues refined to values of 0.59 and 0.41, thus supporting the presence of the two simultaneous conformers.This protein region is relevant since Thr208 and Gly209, participate in the recognition of G3P in the binding site.The large electron density observed in our maps occupies the GP3 binding site, and it was expected that the residues around this site should move to avoid clashing.Interestingly, soaking the crystals with NAD + highly reduces the presence of the large electron density around Cys149 and stabilizes the secondary structure of the alpha helix in only one conformation (Figure 4a,b).All residues recovered the conformation previously reported in other EcG3PDH models [33,34,38,39].

A New Binding Site for Glyceraldehyde 3-Phosphate
In EcGAPDH, residues 148-150 and 208-209 are important for G3P substrate binding, including residues Cys149, and His176, which comprise the catalytic site.After examination of the electron density map, we realized that there was no ligand in the binding pocket.Interestingly, the glyceraldehyde 3-phosphate substrate was found outside the binding site (Figure 5a,b).The density map (2Fo-Fc) showed a complete G3P molecule for monomer A and slightly less electron density for the ligand in monomer B. The phosphate group of G3P interacts with the main chain nitrogen from residues Gly132, Phe135 and Asp136, and also makes a hydrogen bond with the carboxylate oxygen of Asp136 in the NAD + binding domain.The glycerol moiety of G3P is almost completely exposed to the solvent and has no contact with symmetry molecules (Figure 5a).To compare our models with other GAPDH structures in the presence of G3P we made a search in the PDB database using the sequence from the E. coli GAPDH, which resulted in 160 GAPDHs from which 14 structures belong to the E. coli GAPDH.Careful analysis of these models, their ligands and the differences in conformation in the regions reported here was performed.From the 160 GAPDH structures analyzed, 10 contained G3P or an analog, which was present only in the active site of the enzyme.The 14 EcGAPDH models deposited in the PDB were superimposed with the ones obtained here, and we observed that two structures had a sulfate ion in the same place where G3P is bound in our models (PDBs 1s7c and 5za0).
observed that two structures had a sulfate ion in the same place where G3P is bound in our models (PDBs 1s7c and 5za0).
Interestingly, from all EcGAPDHs models deposited in the PDB, only ours exhibit a different conformation of Met267 in monomer A (Figure 5b).This conformation creates a small cleft were the G3P binds, and seems to stabilize G3P binding, in contrast with monomer B where the conformation of this methionine is the same as the one present in all other E. coli models, in this case, less electron density for the ligand is observed.To the best of our knowledge, this is the first GAPDH structure with a large modification in the catalytic Cys149 and the only one with G3P bound to a non-catalytic site.Interestingly, from all EcGAPDHs models deposited in the PDB, only ours exhibit a different conformation of Met267 in monomer A (Figure 5b).This conformation creates a small cleft were the G3P binds, and seems to stabilize G3P binding, in contrast with monomer B where the conformation of this methionine is the same as the one present in all other E. coli models, in this case, less electron density for the ligand is observed.To the best of our knowledge, this is the first GAPDH structure with a large modification in the catalytic Cys149 and the only one with G3P bound to a non-catalytic site.

Trehalose Binding Site
Two of the three GAPDH crystals used for X-ray diffraction were soaked for a second in 25% trehalose as cryo-protectant.Nevertheless, the disaccharide bound readily to the enzyme, clear electron density was observed for this molecule in the two models.The binding site for trehalose seems to display some specificity, as it superimposes entirely to one previously observed [33].There are reports of GAPDH from different organisms binding to other sugars, as this enzyme has been proposed to act as a lectin-like protein [40].However, the site for sugar-binding has been depicted in only a few works [33], whereas binding studies are much more frequent [40,41].

Discussion
Glyceraldehyde 3-phosphate dehydrogenase is one of the most studied enzymes in the last 40 years, and what initially seemed like a simple glycolytic enzyme has been demonstrated to perform crucial roles in metabolism and disease.To date, many of the non-canonical functions of GAPDH are of medical interest because of its involvement in the development of diseases like Alzheimer's and cancer, being part of the oxidative stress response, and also because it can be exploited as a target for vaccine therapeutics [42].GAPDH is one of the most conserved glycolytic enzymes, displaying between 50-60% identity among different species [1].Interestingly, while there are several studies about the structure-function relationship of this enzyme regarding its canonical activity, a direct structural connection for the modulation of its non-glycolytic functions is not yet clearly established.Here, we show how a modification of the catalytic Cys149, presumably glutathione, promotes an unstable conformation of the alpha helix comprised by residues 206 to 214 in the E. coli enzyme.The modified GAPDH was obtained from stressed E. coli cells overexpressing a human protein and grown in minimum media, as explained in the results section.In our model, the modification of Cys149 does not alter the global structure of the GAPDH nor the helical region where it is located.Nevertheless, it significantly affects the structural stability of the alpha helix adjacent to the catalytic site, moving the entire secondary structure (Figure 4a).
Because the NAD + binding domain is close to the catalytic cysteine, NAD + soaking experiments were carried out as described in the methods section.To our surprise, the electron density observed for NAD + was not enough to accommodate the coenzyme, and the refined occupancy gave a value of 0.7.However, the region comprising residues 206-214 was stabilized in a single conformation (Figure 4b).Some positive electron density was still present in the surroundings of Cys149, though it was only appropriate to model two oxygen atoms for a sulfinic acid derivative (Figure 4b).Superposition of the structures from the NAD + soaked crystal and the putative GSH modified EcGAPDH with other reported E. coli GAPDHs (PDB entries 2vyn, 1gae, 1gad, 1dc4, 1dc6, 1dc5, 1s7c, 2vyv, 4mvj, 5za0, 6ioj, 6io6, 6io4) showed that this helix is in one predominant conformation in all the structures.Only model 1dc4, which contains a covalently bound transition state G3P displays a similar conformation of the mentioned alpha helix to the one observed for the Se-modified enzyme.It is important to note that only our model displays the two conformations simultaneously.Residues 210-214, from 1dc4, are in a different conformation, which might be attributed to the bulkier moiety attached to Cys149 in our model.
Previously, it has been shown that post-translational modifications can promote changes in the oligomeric state of GAPDH or to promote aggregation [18,25,43,44].Some modifications modulate the binding partners of GAPDH, and at the same time its participation in different cellular processes.The E. coli and H. sapiens enzymes share approximately 66% in sequence identity, and only one of the 206-214 residues is not conserved between these two enzymes.This region contains Lys214 (H.sapiens numbering, 212 E. coli), a crucial residue for the binding of band three transport protein, which is an ion channel in the membrane of erythrocytes, and necessary for the interaction of GAPDH with the membrane [45].Glutathionylation of GAPDH is a response to oxidative stress in cells that reversibly inactivates the enzyme [19].Thus, the high abundance of GAPDH has been proposed as a protective anti-oxidative mechanism for the cell, which sometimes promotes its translocation to different compartments and provides alternative functions [46,47].These findings contribute with structural evidence demonstrating how oxidation of Cys149 induces local conformational changes that could potentially affect binding to other molecules.
To observe a structural effect in the Se-modified crystals, we performed an NAD + soaking experiment.Surprisingly, the presence of NAD + eliminated the positive electron density near the Cys149.This change also stabilized the secondary structure of the alpha helix, which recovered its commonly observed conformation [33,34,38,39].However, we could not build the NAD + moiety due to its low occupancy (0.7); this data might be in agreement with previous studies in which the affinity towards NAD + is decreased for the oxidized enzyme [48].The presence of NAD + in the drop, however, did not affect the presence of G3P in the unusual binding site, which was also present in the native GAPDH obtained from cells overexpressing the human gene.To the best of our knowledge, there is no previous information regarding a secondary site for the substrate, and we ignore if this may have metabolic or functional implications in this or other systems.The residues interacting with the G3P moiety have not been disclosed as particularly important in any other context, and only in two other cases, these same residues were modeled interacting with a sulfate or formate molecule without any apparent significance [33,49].Therefore, further research is required to investigate the importance of a secondary site and its implications for the GAPDH function.We believe that these results provide information about new remarkable conformational changes in GAPDH that seem to be related to the non-glycolytic functions of E. coli GAPDH and which in turn can also be translated to other systems.

Figure 2 .
Figure 2. Three-dimensional structure of GAPDH.(a) A homodimer was crystallized in the asymmetric unit (darker colors), a tetramer is formed by symmetry molecules (light colors).Each GAPDH monomer is composed of two domains, the NAD + binding domain (residues 1-148 and 312-330) is shown in cyan and orange on individual monomers, whereas the catalytic domain (residues 149-311) is shown in blue and red.(b) The NAD + (in sticks) binding site and the catalytic site (yellow) are adjacent to the helix that changes its conformation (shown in green only in the left monomers).NAD + was superimposed in this model using chain C from PDB 4mvj (b, bottom left).

Figure 2 .
Figure 2. Three-dimensional structure of GAPDH.(a) A homodimer was crystallized in the asymmetric unit (darker colors), a tetramer is formed by symmetry molecules (light colors).Each GAPDH monomer is composed of two domains, the NAD + binding domain (residues 1-148 and 312-330) is shown in cyan and orange on individual monomers, whereas the catalytic domain (residues 149-311) is shown in blue and red.(b) The NAD + (in sticks) binding site and the catalytic site (yellow) are adjacent to the helix that changes its conformation (shown in green only in the left monomers).NAD + was superimposed in this model using chain C from PDB 4mvj (b, bottom left).

Figure 3 .
Figure 3. Electron density in the catalytic cysteine.(a) The Se-modified GAPDH displayed a large electron density surrounding the catalytic Cys149, the images show a mesh in cyan the Fo-Fc map (σ level 3.0) and a dark blue mesh for the 2Fo-Fc map (σ level 1.0) after final model refinement.(b) A likely oxidizer of the GAPDH enzyme is glutathione; nevertheless, it was removed from the final model due to clashes with the protein.

Figure 3 .
Figure 3. Electron density in the catalytic cysteine.(a) The Se-modified GAPDH displayed a large electron density surrounding the catalytic Cys149, the images show a mesh in cyan the Fo-Fc map (σ level 3.0) and a dark blue mesh for the 2Fo-Fc map (σ level 1.0) after final model refinement.(b) A likely oxidizer of the GAPDH enzyme is glutathione; nevertheless, it was removed from the final model due to clashes with the protein.

Figure 3 .
Figure 3. Electron density in the catalytic cysteine.(a) The Se-modified GAPDH displayed a large electron density surrounding the catalytic Cys149, the images show a mesh in cyan the Fo-Fc map (σ level 3.0) and a dark blue mesh for the 2Fo-Fc map (σ level 1.0) after final model refinement.(b) A likely oxidizer of the GAPDH enzyme is glutathione; nevertheless, it was removed from the final model due to clashes with the protein.

Figure 4 .
Figure 4. Double conformation of residues 206-214 in the catalytic site of GAPDH.(a) A double conformation of residues 206-214 is observed in the GAPDH structure with Se-methionines (cyan and purple).This structure contains a large electron density attached to the catalytic cysteine Cys149 (Fo-Fc map in green).(b).One conformation is stabilized for residues 206-214.When the crystals are soaked in a NAD + solution (purple), the electron density for the surroundings of Cys149 disappears (Fo-Fc map green mesh).

Figure 4 .
Figure 4. Double conformation of residues 206-214 in the catalytic site of GAPDH.(a) A double conformation of residues 206-214 is observed in the GAPDH structure with Se-methionines (cyan and purple).This structure contains a large electron density attached to the catalytic cysteine Cys149 (Fo-Fc map in green).(b).One conformation is stabilized for residues 206-214.When the crystals are soaked in a NAD + solution (purple), the electron density for the surroundings of Cys149 disappears (Fo-Fc map green mesh).

Figure 5 .
Figure 5. New binding site for glyceraldehyde 3 phosphate (G3P).(a) A shallow binding site on the GAPDH surface was found for the GAPDH substrate, the glycerol moiety of G3P (ball representation)

Figure 5 .
Figure 5. New binding site for glyceraldehyde 3 phosphate (G3P).(a) A shallow binding site on the GAPDH surface was found for the GAPDH substrate, the glycerol moiety of G3P (ball representation) is completely exposed to solvent, and no symmetry related molecules are near to the site.(b) G3P is anchored to the enzyme mainly through main chain interactions, and only one residue Glu266 interacts indirectly with its side chain and a solvent molecule with one phosphate oxygen of G3P.

Table 1 .
Data collection and refinement statistics.